AI-900 retired on June 30, 2026. If you’re only finding this now, the decision has already been made for you: AI-901 is the only route left to “Azure AI Fundamentals,” and it’s a different kind of exam.
Here’s exactly what changed and how to prepare for it.
The Certification Didn’t Retire, Just the Old Exam
“Microsoft Certified: Azure AI Fundamentals” as a credential is still here. Only the path to earn it changed.
AI-900 is gone. AI-901 is now the only way to earn the same certification. If you already passed AI-900, nothing changes. It stays on your transcript and doesn’t expire.
AI-901 Is Not a Renamed AI-900
This is the part that catches people off guard.
AI-900 was a conceptual exam. Five descriptive domains and no coding required, open to technical and non-technical candidates alike:
| Domain | Weight |
|---|---|
| AI workloads and considerations | 15–20% |
| Machine learning principles on Azure | 15–20% |
| Computer vision workloads on Azure | 15–20% |
| NLP workloads on Azure | 15–20% |
| Generative AI workloads on Azure | 20–25% |
AI-901 is implementation-heavy. Two domains, and more than half the exam is hands-on work in Azure:
| Domain | Weight |
|---|---|
| AI concepts and capabilities | 40–45% |
| Implement AI solutions with Microsoft Foundry | 55–60% |
Microsoft’s AI-901 exam page is explicit: candidates need Python coding knowledge, familiarity with REST APIs, SDKs, and CLIs. The exam tests whether you can deploy and interact with models in Azure AI Foundry, not just describe what they do.
Passing score for both: 700 out of 1000.
What This Means for You Now
If you already passed AI-900. You’re done. The credential is on your transcript and stays valid, nothing to redo. If you want to go deeper, a role-based Azure cert is the natural next step; the cloud certification roadmap shows where each one fits.
If you were mid-prep for AI-900. Your conceptual foundation still counts. AI-901’s first domain (AI concepts and capabilities, 40–45%) overlaps heavily with what you already studied. The new work is the implementation half: getting hands-on in Azure AI Foundry with Python, REST APIs, and the SDKs.
If you’re starting from scratch. Go straight at AI-901. There’s no shortcut left, and it’s a more relevant credential anyway. It’s built around the tooling Azure AI teams use: deploying models, building agents, working with multimodal inputs. The skills map directly to current AI roles.
How to Prepare for AI-901
The implementation domain is the whole game. Fifty-five to sixty percent of AI-901 is “Implement AI solutions with Microsoft Foundry.” If you come from AI-900’s descriptive world, this is where the gap is. Get into Azure AI Foundry and deploy something: spin up a model, call it from the SDK, wire up a simple agent. Reading about it isn’t enough.
Know the Foundry stack, not just the concepts. Questions are scenario-based: “which service should you use to…” rather than “what is…”. Know when to reach for Azure AI Foundry vs Azure OpenAI Service vs the broader Azure AI catalog, and how the SDKs, REST APIs, and CLIs fit together.
Don’t skip responsible AI. Microsoft’s six principles (fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability) still appear throughout, tested through deployment scenarios rather than definitions. A question describes a design decision and asks which principle is at stake.
Pass score: 700 out of 1000. Microsoft’s official AI-901 study guide covers every objective.
Pass-IT has AI-901 adaptive practice questions grounded in the current exam guide, including the hands-on Foundry scenarios. Per-domain tracking shows exactly where you’re losing points before you book. Free to start, no card for your first session.
